Research ArticleCancer

Hypoxic cancer–associated fibroblasts increase NCBP2-AS2/HIAR to promote endothelial sprouting through enhanced VEGF signaling

See allHide authors and affiliations

Science Signaling  05 Feb 2019:
Vol. 12, Issue 567, eaan8247
DOI: 10.1126/scisignal.aan8247

The secret(ing) life of the tumor stroma

The hypoxia that develops in solid tumors leads to the formation of a dysfunctional vasculature that enables cancer cell survival and also prevents efficient chemotherapeutic penetration. Cancer-associated fibroblasts (CAFs) in the tumor stroma release important angiogenic signals; however, the proteins secreted by these cells have generally been characterized at the genome level. Kugeratski et al. analyzed the proteome and secretome of CAFs and found that hypoxia altered the secretome of these cells. The authors identified a previously uncharacterized protein that they renamed HIAR. HIAR was increased in abundance in hypoxic CAFs, promoted the release of the pro-angiogenic factor VEGF from CAFs, and induced VEGF-dependent signaling in endothelial cells. These results provide an overview of the CAF secretome at the protein level and a new potential target for anti-angiogenic therapy.

Abstract

Intratumoral hypoxia causes the formation of dysfunctional blood vessels, which contribute to tumor metastasis and reduce the efficacy of therapeutic treatments. Blood vessels are embedded in the tumor stroma of which cancer-associated fibroblasts (CAFs) constitute a prominent cellular component. We found that hypoxic human mammary CAFs promoted angiogenesis in CAF-endothelial cell cocultures in vitro. Mass spectrometry–based proteomic analysis of the CAF secretome unraveled that hypoxic CAFs contributed to blood vessel abnormalities by altering their secretion of various pro- and anti-angiogenic factors. Hypoxia induced pronounced remodeling of the CAF proteome, including proteins that have not been previously related to this process. Among those, the uncharacterized protein NCBP2-AS2 that we renamed HIAR (hypoxia-induced angiogenesis regulator) was the protein most increased in abundance in hypoxic CAFs. Silencing of HIAR abrogated the pro-angiogenic and pro-migratory function of hypoxic CAFs by decreasing secretion of the pro-angiogenic factor VEGFA and consequently reducing VEGF/VEGFR downstream signaling in the endothelial cells. Our study has identified a regulator of angiogenesis and provides a map of hypoxia-induced molecular alterations in mammary CAFs.

INTRODUCTION

Intratumoral hypoxia favors tumor aggressiveness and is associated with risk of metastasis, limited response to therapies, and poor clinical outcome. The tumor vasculature plays key roles in these processes. The uncontrolled growth of the cancer cells results in the rapid consumption of oxygen released by the adjacent vasculature, leading to the formation of intratumoral hypoxic regions. To counteract the lack of oxygen, hypoxic cancer and stromal cells secrete factors that stimulate the recruitment of blood vessels, a process that is called angiogenesis (1). However, rather than promoting the formation of a functional vasculature, excessive amounts of those factors result in the recruitment of blood vessels that are leaky and nonfunctional. This phenomenon, in turn, leads to the expansion of hypoxic areas and the formation of a positive feedback loop between hypoxia and nonfunctional blood vessels that is detrimental and challenging to disrupt (24). So far, most of the effort has been devoted to targeting the vascular endothelial growth factor (VEGF) pathway (5), which is a potent inducer of blood vessel growth and is transcriptionally regulated by the hypoxia-inducible factor 1α (HIF1α) (6). However, VEGF also controls the growth of blood vessels in normal tissues (7); therefore, therapies that block VEGF signaling result in unwanted adverse effects (8). The identification of mechanisms that regulate blood vessel recruitment in hypoxic tumors may provide useful hints to develop strategies to block this feedback loop without interfering with the normal vasculature. Tumor blood vessels are typically embedded within the stroma, where cancer-associated fibroblasts (CAFs) are an abundant cell population (9). CAFs can originate from activation of resident normal fibroblasts within the tumor stroma and secrete factors that alter the composition and physical properties of the extracellular matrix (ECM) and signal to adjacent cells (1012). CAFs play crucial roles in the pathogenesis of cancer, including the recruitment of blood vessels within the tumor (13, 14). How CAFs respond to oxygen deprivation is largely unknown and rather controversial. Intratumoral hypoxia induces CAF activation and HIF1α-activated fibroblasts cotransplanted in a MDA-MB-231 breast cancer xenograft are pro-tumorigenic and pro-metastatic (15). Conversely, tumorigenesis in a PyMT breast cancer model is accelerated by fibroblast-specific depletion of Hif1α before tumor onset, but not by depletion of prolyl-hydroxylase 2 (Phd2), which stabilizes HIF1α (16, 17). Furthermore, decreasing PHD2 levels in CAFs reduces their activated phenotype and their pro-metastatic activity in a 4T1 breast cancer orthotopic model (18). A better understanding of the molecular mechanisms underpinning the response of CAFs to hypoxia could bring useful insights on the role of these cells in hypoxic tumors.

Hypoxia-induced signaling has been mostly studied at the genome-wide gene expression level, and there is little information on how it affects the proteome of the cells. Modern mass spectrometry (MS) technology and robust platforms for data analysis allow in-depth and highly accurate quantitative analysis of cellular and subcellular proteomes (1921), including those shaped by hypoxia (22, 23). Proteome-wide analysis of breast cancer cell–derived conditioned medium (CM) has shown that high levels of lysyl oxidase (LOX) are induced by hypoxia and associated with bone metastasis, leading to the discovery of a role for LOX in bone metastasis formation (23).

Here, we have used three-dimensional (3D) cocultures of human primary endothelial cells (ECs) with mammary CAFs and established that the secretome of hypoxic CAFs promotes EC sprouting, which is one of the initial steps of angiogenesis (24). We have identified proteomic alterations induced by hypoxia in CAFs using high-resolution MS and stable isotope labeling with amino acids in cell culture (SILAC) (25). We found hundreds of regulated molecules that may support hypoxic CAFs to drive angiogenesis and discovered a regulatory mechanism of VEGF signaling through which hypoxic but not normoxic CAFs control endothelial sprouting.

RESULTS

The secretome of hypoxic CAFs promotes endothelial sprouting

To determine how the hypoxic environment influences the ability of CAFs to regulate angiogenesis, we used an established 3D coculture model in which collagen IV–coated beads are covered with ECs and embedded into fibrin gel. CAFs are then plated on top of the solidified gel and secrete factors that influence the sprouting and invasion of the ECs through the 3D matrix (Fig. 1A) (26). We used human umbilical vein ECs because they are a well-characterized model to study endothelial functions and angiogenesis (27, 28). For fibroblasts, we used two human mammary CAF cell lines: cancer cell–induced CAFs (cCAFs) that we and others have previously characterized for their pro-angiogenic function in vitro and in vivo (13, 29) and patient-derived CAFs (pCAFs), which were isolated from a triple-negative invasive ductal carcinoma. The coculture was exposed to either normoxia or hypoxia for 72 hours. Endothelial sprouting was enhanced by hypoxia in both cocultures compared with their normoxic counterpart, as measured by the total number and length of EC sprouts and their migration area surrounding the bead (Fig. 1B). To determine whether the factors secreted by hypoxic CAFs were the major contributor of this phenotype, normoxic ECs embedded in fibrin gel were treated with CM derived from cCAFs or pCAFs cultured under hypoxia or normoxia for 72 hours (Fig. 1A). CAFs readily responded to hypoxia, because HIF1α was stabilized and accumulated into the nucleus after 4 hours of oxygen deprivation (fig. S1, A and B). Moreover, after 72 hours of hypoxia, CAFs were still proliferative and not apoptotic, as measured by 5-ethynyl-2′-deoxyuridine (EdU) incorporation (fig. S1C) and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay (fig. S1D). Using this setup, where there are no CAFs in the 3D culture and ECs are stimulated by CAF-derived CM, we observed a significant increase in sprouting when ECs were treated with the hypoxic CM (Fig. 1C). Hence, hypoxic CAFs secrete factors that promote endothelial sprouting.

Fig. 1 The secretome of hypoxic CAFs induces endothelial sprouting angiogenesis.

(A) Scheme of the 3D fibrin-based sprouting assay for ECs cocultured with CAFs or alone and treated with CAF-derived CM. (B) Representative images and quantification of the sprouting of ECs cocultured with CAFs under normoxia and hypoxia for 72 hours. Scale bar, 50 μm. N = beads assessed in three biological replicates; cCAF normoxia, n = 120; cCAF hypoxia, n = 135; pCAF normoxia, n = 128; pCAF hypoxia, n = 134. (C) Representative image and quantification of the sprouting of ECs treated with CM derived from CAFs cultured under normoxia or hypoxia for 72 hours. Scale bar, 50 μm. N = beads assessed in three biological replicates; cCAF normoxia, n = 201; cCAF hypoxia, n = 195; pCAF normoxia, n = 112; pCAF hypoxia, n = 135. (D) Reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis showing VEGFA transcription in cCAFs and pCAFs under hypoxic condition (normalized to the VEGFA levels in normoxic condition. VEGFA normoxia = 1 = dashed line). VEGFA mRNA levels were normalized to TBP. N = three biological replicates, one-sample t test. (E) Enzyme-linked immunosorbent assay (ELISA) showing VEGFA levels in the CM from cCAFs and pCAFs. N = two biological replicates. (F) Quantification of the sprouting of ECs treated with the CM from normoxic and hypoxic CAFs in the presence of bevacizumab (1 μg/ml) or immunoglobulin K (IgGK) (control). N = beads assessed in three biological replicates; CM normoxia + IgGK, n = 178; CM hypoxia + IgGK, n = 161; CM normoxia + bevacizumab, n = 176; CM hypoxia + bevacizumab, n = 179. *P < 0.05, **P < 0.01, ****P < 0.0001, t test.

Hypoxia increases the expression of the potent pro-angiogenic factor VEGFA by activating the HIF pathway. As expected, hypoxia enhanced the mRNA levels and secretion of VEGFA in both cCAFs and pCAFs (Fig. 1, D and E). To assess whether the pro-angiogenic effect of the CM from hypoxic CAFs was due to increased VEGFA levels, we blocked VEGFA/VEGFR signaling with the anti-VEGFA antibody bevacizumab. The addition of bevacizumab to the CM from hypoxic and normoxic CAFs reduced the sprouting of ECs by 60%, indicating that VEGF is a major contributor of the pro-angiogenic function of CAFs under either condition (Fig. 1F). This finding was confirmed by the reduced sprouting observed upon HIF1α silencing in hypoxic CAFs (fig. S1E). Factors other than VEGF may also contribute to the pro-angiogenic effect of the hypoxic CM, because even in the presence of bevacizumab, the CM from hypoxic CAFs enhanced endothelial sprouting (Fig. 1F).

Hypoxia remodels the proteome and secretome of CAFs to promote angiogenesis

To determine the molecular events underpinning the pro-angiogenic function of hypoxic CAFs, we used MS-based proteomics to compare the proteome of cCAFs and pCAFs cultured under normoxia or hypoxia for 72 hours, using SILAC for accurate protein quantification. We additionally analyzed the secretome of these cells to map the composition of the pro-angiogenic secretome of hypoxic CAFs. We measured the total cell proteome by fractionating with strong anion exchange chromatography tryptic peptides obtained from SDS-extracted proteins (30). We separated the CM into two fractions using multistep centrifugation: a fraction containing soluble factors (soluble) and a fraction containing extracellular vesicles (EVs). All samples were measured on an Orbitrap mass spectrometer followed by analysis with the MaxQuant computational platform (Fig. 2A) (31). Overall, 3513 proteins were robustly quantified in the cell proteome (in at least two of three biological replicates in both cCAFs and pCAFs), 681 proteins in the soluble secretome, and 165 proteins in the EV fraction (data file S1). Because we found that replicates strongly correlated with each other (fig. S1F), for each protein, we averaged the hypoxia/normoxia SILAC ratio between replicates; proteins were defined as hypoxia-regulated based on the SD of their ratio from the mean SILAC ratio calculated for the entire dataset. We classified regulated proteins as strongly regulated when the hypoxia/normoxia ratio was greater than two SDs from the mean or as modestly regulated when the hypoxia/normoxia ratio was between one and two SDs. The latter group contained well-established HIF1α target genes, such as CXCL12, LOX-like 2 (LOXL2), procollagen-lysine,2-oxoglutarate 5-dioxygenase 2 (PLOD2), and prolyl 4-hydroxylase subunit α-1 (P4HA1), among others (data files S1 and S2) (3234), supporting the choice of this threshold to identify hypoxia-controlled proteins in our study. We required proteins to be regulated in both cCAFs and pCAFs. With these criteria, we identified 124 up-regulated and 221 down-regulated proteins by hypoxia in the intracellular proteome and 27 up-regulated and 34 down-regulated proteins in the secretomes (fig. S2A). The physical and functional interactions between these proteins and the enriched functional categories showed that a vast proportion mirrored the metabolic reprogramming required for cellular adaptation to reduced oxygen availability (Fig. 2B). Hypoxia increased the levels of proteins involved in glycolysis, while it decreased the levels of those involved in oxidative phosphorylation. In addition, this analysis revealed a significant regulation of extracellular proteins, such as those located in the ECM or in exosomes and EVs (Fig. 2B), thus underlining the impact that hypoxia exerts on the ability of CAFs to alter the surrounding environment. In addition, we pinpointed the angiogenesis-regulating factors that were secreted by hypoxic CAFs. These included stanniocalcin 1 (STC1) (35), LOXL2 (32, 36), CXCL12 (14), annexin 1 and 2 (37, 38), and the endotrophin fragment of collagen 6-α (39), all of which were up-regulated by hypoxia (fig. S2, A and B). We confirmed VEGFA, LOXL2, and STC1 up-regulation by RT-qPCR analysis (fig. S2C). Conversely, collagen fragments with anti-angiogenic properties, such as vastatin (COL8A1) (40), arresten (COL4A2) (41), and canstatin (COL4A1) (42), were down-regulated (fig. S2, A and B).

Fig. 2 Hypoxia-induced remodeling of the proteome and secretome of mammary CAFs.

(A) SILAC-based workflow used to characterize the total proteome and secretome of hypoxic cCAFs and pCAFs. (B) Physical and functional relationship among the hypoxia-regulated proteins in the intracellular proteome and secretome, as determined by STRING analysis and visualized with Cytoscape. Gene ontology (GO) categories and KEGG pathways enriched in the subset of proteins up-regulated and down-regulated by hypoxia are indicated on the right. FASP, filter-aided sample preparation; SAX, strong anion exchange. Results from n = three biological replicates for both cCAF and pCAF.

NCBP2-AS2 is a hypoxia-induced factor in CAFs

Next, we exploited our data to identify mechanisms that underpin hypoxic CAF functions. We compared our hypoxia-induced proteomic signature with known HIF1α-responsive genes and publicly available hypoxia-related gene expression signatures (fig. S3A and data file S2). More than 75% of our signature (300 proteins) was not identified as hypoxia-regulated at the gene expression level by any of the analyzed datasets. This result suggests that the extent of gene expression regulation was below the statistical threshold or that these proteins are regulated by hypoxia either posttranscriptionally or specifically in CAFs. Supporting the hypothesis of posttranscriptional regulation, hypoxia down-regulated pentraxin-related protein 3 (PTX3) and carboxypeptidase A4 (CPA4) at the protein, but not at the gene expression, level (fig. S3, B and C). Moreover, we identified 134 mitochondrial proteins with decreased levels in hypoxic CAFs (fig. S3A and data file S2), indicating that hypoxia may also induce extensive mitochondrial damage in CAFs (22, 43).

Within our signature, there were proteins concordantly regulated in both the proteome and secretome: COL1A1, COL6A2, COL14A1, LRRC17, NCBP2-AS2, SDC2, and VCAM1 (fig. S3A and data file S2). These proteins could be important because they may have a strong impact not only on CAF functions but also as paracrine factors on other cell types; moreover, they could potentially be used as biomarkers for hypoxia. Of these, NCBP2-AS2 was strongly up-regulated by hypoxia in the proteome and soluble secretome of cCAFs and pCAFs (Fig. 3, A and B, and data file S2). Fewer peptides of NCBP2-AS2 were identified in the secretome than in the proteome (data file S2), suggesting that NCBP2-AS2 is predominantly intracellular but is also released as a soluble factor. Increased levels of NCBP2-AS2 in total CAF lysate were confirmed using a targeted MS-based approach (fig. S4, A to C). Conversely, we did not detect a corresponding increase in NCBP2-AS2 at the mRNA level in hypoxic CAFs (Fig. 3C and fig. S5, A and B) or upon treatment with the HIF1α-stabilizing prolyl hydroxylase domain–containing protein inhibitor dimethyloxaloylglycine (DMOG) (Fig. 3D). Conversely, NCBP2-AS2 translation increased with hypoxia. Ribosomal profiling analysis showed that normoxic and hypoxic CAFs had similar levels of total protein translation (Fig. 3E). Furthermore, the amount of NCBP2-AS2 mRNA associated with polysomes, which are the actively translating subset of ribosomes, was significantly higher in hypoxic than in normoxic CAFs (Fig. 3F). We confirmed an increase in NCBP2-AS2 translation upon stimulation with hypoxia also using a pulsed-SILAC approach coupled to targeted MS proteomics (fig. S6A). Thus, our findings indicate that hypoxia induces NCBP2-AS2 posttranscriptionally by enhancing its translation.

Fig. 3 Hypoxic CAFs increase NCBP2-AS2 protein levels through enhanced translation.

(A and B) Scatter plot of the averaged SILAC ratio, hypoxia/normoxia, measured by MS in the pCAF and cCAF proteome (A) and secretome (B) highlighting the induction of NCBP2-AS2 by hypoxia in both cell types. Results from n = 3 biological replicates. (C) RT-qPCR showing mRNA levels of NCBP2-AS2 and VEGFA normalized to TBP in CAFs upon exposure to hypoxia for 72 hours (normalized to the levels in CAFs in normoxia. Normoxia = 1 = dashed line). VEGFA was used as a positive control. N = three biological replicates. One-sample t test. (D) RT-qPCR showing the mRNA levels of NCBP2-AS2 and VEGFA upon treatment with 1 mM DMOG for 24 hours [normalized to the levels in dimethyl sulfoxide (DMSO) control-treated cells. DMSO = 1 = dashed line]. The mRNA levels were normalized to TBP. VEGFA was used as a positive control. N = three biological replicates. One-sample t test. (E) Polysomal profiling of cCAFs exposed to hypoxia and normoxia for 24 hours. Representative of three biological replicates. (F) RT-qPCR analysis for NCBP2-AS2 and actin mRNAs, which was used as control, on the nontranslating and polysomal ribosomal fractions. N = three biological replicates. *P < 0.05, **P < 0.01, t test. NS, not significant.

Hypoxic CAFs promote angiogenesis through NCBP2-AS2

To determine whether NCBP2-AS2 contributed to the pro-angiogenic function of hypoxic CAFs, we silenced NCBP2-AS2 in CAFs and assessed the pro-angiogenic function of their CM. ECs treated with the CM from hypoxic CAFs with NCBP2-AS2 silencing sprouted substantially less than those treated with the CM from control CAFs. Conversely, the CM from normoxic CAFs control or those with NCBP2-AS2 silencing promoted endothelial sprouting to a similar extent (Fig. 4, A to D, and fig. S6B). Of note, the sprouting induced by normoxic and hypoxic CAFs with NCBP2-AS2 silencing was comparable. Because we did not detect a significant increase in sprouting when the recombinant purified glutathione S-transferase (GST)–tagged NCBP2-AS2 was added to the medium of ECs cultured in 3D fibrin gel (fig. S7A), NCBP2-AS2 is likely to control endothelial sprouting by regulating intracellular CAF functions rather than as a paracrine factor. Next, we confirmed that NCBP2-AS2 supports the pro-angiogenic function of hypoxic CAFs also in vivo in a subcutaneous sponge assay typically used to investigate angiogenesis in vivo (44). In this assay, a sponge is subcutaneously implanted to initiate an inflammatory reaction that induces EC infiltration. This leads to the formation of a functional capillary network, whose extension is determined by the factors or CM contained in the sponge. The vessel density in sponges that were embedded with the CM from CAFs with NCBP2-AS2 silencing was significantly lower than in those containing the CM from control CAFs (Fig. 4, E and F).

Fig. 4 The CM from hypoxic CAFs silenced for NCBP2-AS2 show a decreased pro-angiogenic activity in vitro and in vivo.

(A and C) Sprouting quantification of ECs treated with CM from hypoxic and normoxic cCAFs (A) or pCAFs (C) transfected with siCTL or siNCBP2-AS2 [two different small interfering RNAs (siRNAs)]. N = beads assessed in three biological replicates; cCAF hypoxic CM: siCTL, n = 129; siNCBP2-AS2 #1, n = 124; and siNCBP2-AS2 #2, n = 116. Normoxic CM: siCTL, n = 121; siNCBP2-AS2 #1, n = 131; and siNCBP2-AS2 #2, n = 122. pCAF hypoxic CM: siCTL, n = 119; siNCBP2-AS2 #1, n = 107; and siNCBP2-AS2 #2, n = 97. Normoxic CM: siCTL, n = 111 beads; siNCBP2-AS2 #1, n = 127; and siNCBP2-AS2 #2, n = 104. (B and D) RT-qPCR showing the remaining mRNA levels of NCBP2-AS2 normalized to TBP in (B) cCAFs and (D) pCAFs used to produce CM. (E) Vascular density quantification, as measured by Meca32+ and Ng2+ staining of the blood vessels (y axis), of subcutaneous sponges, which were embedded with CM derived from hypoxic CAFs transfected with siCTL or siNCBP2-AS2. N = 1/2 sponges derived from six mice per group; t test. (F) RT-qPCR showing the remaining mRNA levels of NCBP2-AS2 normalized to TBP in CAFs producing the CM used in the sponge assay in (E). *P < 0.05, ***P < 0.001, ****P < 0.0001.

Because sprouting angiogenesis is regulated by a combination of EC proliferation and migration, we investigated how NCBP2-AS2 silencing affected these functions. When treated with the CM from CAFs with or without NCBP2-AS2 silencing, EC proliferation in 3D and 2D culture was comparable, as measured by EdU incorporation (Fig. 5, A to C, and fig. S7, B and C). Next, we established that the CM from hypoxic CAFs with NCBP2-AS2 silencing had reduced pro-migratory properties compared with the CM from hypoxic CAFs control. Wound closure in an endothelial monolayer was slower in ECs treated with CM from NCBP2-AS2–silenced hypoxic CAFs. Conversely, silencing of NCBP2-AS2 in normoxic CAFs had minor impact on the pro-migratory function of their secretome (Fig. 5, D to G). To confirm the role of NCBP2-AS2 in cell migration, we evaluated the ability of ECs to form lamellipodia, because pro-angiogenic factors such as VEGF induce the formation of cell protrusions (45). To monitor lamellipodia, we performed immunostaining for cortactin, which is a cortical actin remodeling protein and a major mediator of lamellipodia formation (46). The presence or absence of NCBP2-AS2 in hypoxic CAFs did not affect the total perimeter of ECs, but when treated with the CM from hypoxic CAFs with NCBP2-AS2 silencing, the ECs had significantly fewer cortactin-enriched protrusions (Fig. 5, H to J). These data show that NCBP2-AS2 in CAFs is required to support the pro-angiogenic activity of their secretome.

Fig. 5 ECs treated with the CM from hypoxic CAFs silenced for NCBP2-AS2 have a decreased migratory capacity.

(A) Proliferation of ECs cultured in 2D and treated with the CM from normoxic and hypoxic CAFs transfected with siCTL or siNCBP2-AS2. N = 45 fields assessed per condition from three biological replicates. (B) Proliferation of ECs cultured in 3D fibrin gel and treated with the CM from normoxic and hypoxic CAFs transfected with siCTL or siNCBP2-AS2. N = beads assessed in three biological replicates. Normoxic CM: siCTL, n = 124; siNCBP2-AS2 #1, n = 132; and siNCBP2-AS2 #2, n = 136. Hypoxic CM: siCTL, n = 127; siNCBP2-AS2 #1, n = 120; and siNCBP2-AS2 #2, n = 134. (C) RT-qPCR showing remaining mRNA levels of NCBP2-AS2 normalized to TBP. (D) RT-qPCR showing remaining mRNA levels of NCBP2-AS2 normalized to TBP. (E and F) Scratch wound migration assay using ECs treated with the CM from hypoxic (E) or normoxic (F) cCAFs transfected with siCTL or siNCBP2-AS2. N = two biological replicates. (G) Relative wound density measured at 12 hours in the two independent replicate experiments [dashed lines in (E) and (F)]. (H) Representative images of lamellipodia formation in ECs treated with the CM from cCAFs silenced or not for NCBP2-AS2. (I) Quantification of cortactin at the cell membrane. (J) Quantification of the total cell perimeter. N = cells measured in three biological replicates. siCTL, n = 111; siNCBP2-AS2 #1, n = 126; and siNCBP2-AS2 #2, n = 125. Scale bar, 10 μm; scale bar in zoomed inset, 5 μm. P value based on Kruskal-Wallis test corrected for multiple testing using Dunn’s test; t test was used for the other panels. **P < 0.01, ****P < 0.0001.

NCBP2-AS2 regulates VEGFA expression

To understand how NCBP2-AS2 controls the pro-sprouting function of the CAF secretome, we used SILAC MS proteomics to identify molecular alterations induced upon silencing of NCBP2-AS2 in hypoxic CAFs. The knockdown of NCBP2-AS2 did not significantly alter the intracellular proteome of the CAFs (Fig. 6A and data file S3). Conversely, the levels of VEGFA in the secretome of CAFs silenced for NCBP2-AS2 were down-regulated (Fig. 6B and data file S4), as also confirmed by ELISA (Fig. 6, C and D). RT-qPCR analysis showed that VEGFA down-regulation also occurred at the transcriptional level (Fig. 6, E and F). VEGF levels in CAFs were not regulated in an autocrine manner by secreted NCBP2-AS2, because the amount of VEGFA in the secretome of hypoxic CAFs silenced for NCBP2-AS2 could not be rescued upon treatment with CM from hypoxic control CAFs (fig. S7, D to F). Next, we investigated how other hypoxia-regulated pro- and anti-angiogenic factors were affected by NCBP2-AS2 silencing in CAFs. The levels of the pro-angiogenic protein STC1 decreased upon NCBP2-AS2 silencing, although at a lower extent compared to VEGFA, whereas the anti-angiogenic factor COL4A2 was modestly increased upon NCBP2-AS2 silencing (Fig. 6G). Thus, STC1 and COL4A2 may also contribute to the decreased pro-angiogenic function of hypoxic CAFs. Together, these results show that VEGFA could account for most of the reduced pro-angiogenic effect of the secretome of CAFs with NCBP2-AS2 silencing.

Fig. 6 NCBP2-AS2 regulates the levels of CAF-derived VEGFA.

(A and B) Volcano plot showing the SILAC ratio of the proteins quantified in the total proteome (A) and secretome (B) of hypoxic cCAFs transfected with siCTL and siNCBP2-AS2. The dashed lines show the regulation cutoff of at least onefold change and P < 0.05. Results from three biological replicates. (C) ELISA of VEGFA in the CM from hypoxic CAFs silenced or not for NCBP2-AS2 and normoxic CAFs siCTL. One-tailed t test. N = three or four biological replicates. (D) RT-qPCR showing remaining mRNA levels of NCBP2-AS2 normalized to TBP in cells used in (C). (E) RT-qPCR for VEGFA in hypoxic cCAFs and pCAFs transfected with siCTL or siNCBP2-AS2. The mRNA levels were normalized to TBP. One-sample t test. N = three biological replicates. (F) RT-qPCR for NCBP2-AS2 showing the silencing efficiency in cells used in (E). NCBP2-AS2 mRNA levels were normalized to TBP. N = three biological replicates. (G) Influence of NCBP2-AS2 silencing on the angiogenic secretome of hypoxic CAFs. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

We reasoned that reduced amounts of VEGFA in the CM from NCBP2-AS2–silenced hypoxic CAFs caused the reduction in sprouting observed in ECs treated with this CM. Defective sprouting in ECs was rescued by supplementing the CM from hypoxic NCBP2-AS2–silenced CAFs with VEGFA (Fig. 7, A to C), thus confirming our hypothesis. Likewise, the addition of bevacizumab in the CM from hypoxic CAFs with NCBP2-AS2 silencing did not have an additive inhibitory effect on endothelial sprouting (Fig. 7, D to F). Moreover, ECs treated with hypoxic CAF-derived CM had activated VEGFR2 and canonical downstream signaling, as shown by phosphorylation of the regulatory site of phospholipase Cγ (PLCγ) (Tyr783), AKT (Ser473), and extracelullar signal–regulated kinase (ERK) 1/2 (Thr202/Tyr204). Conversely, the extent of VEGFR2 activation and the downstream signaling were reduced in ECs treated with CM from hypoxic CAFs silenced for NCBP2-AS2 (Fig. 7, G and H, and fig. S8). Thus, we show that NCBP2-AS2 controls the pro-angiogenic function of CAF secretome by regulating VEGFA levels and that this, in turn, controls VEGF-dependent sprouting in ECs (Fig. 7I).

Fig. 7 NCBP2-AS2 regulates EC sprouting through VEGFA signaling.

(A) Sprouting quantification of ECs treated with the CM from hypoxic cCAFs transfected with siCTL or siNCBP2-AS2 (two different siRNAs) with or without VEGFA. N = beads assessed in three biological replicates. CM without VEGFA: siCTL, n = 149; siNCBP2-AS2 #1, n = 128; and siNCBP2-AS2 #2, n = 110. CM with VEGFA: siCTL, n = 132; siNCBP2-AS2 #1, n = 130; and siNCBP2-AS2 #2, n = 126. (B) Sprouting quantification of ECs treated with the CM from hypoxic pCAFs transfected with siCTL or siNCBP2-AS2 (two different siRNAs) with or without VEGFA. N = beads assessed in three biological replicates. CM without VEGFA: siCTL, n = 162 beads; siNCBP2-AS2 #1, n = 132; and siNCBP2-AS2 #2, n = 124. CM with VEGFA: siCTL, n = 127; siNCBP2-AS2 #1, n = 133; and siNCBP2-AS2 #2, n = 136. (C) RT-qPCR showing remaining mRNA levels of NCBP2-AS2 normalized to TBP. (D) Sprouting quantification of ECs treated with the CM from hypoxic cCAFs transfected with siCTL or siNCBP2-AS2 (two different siRNAs) in the presence of bevacizumab or IgGK (control). N = beads assessed in three biological replicates. CM with IgG: siCTL, n = 135; siNCBP2-AS2 #1, n = 129; and siNCBP2-AS2 #2, n = 107. CM with bevacizumab: siCTL, n = 127; siNCBP2-AS2 #1, n = 147; and siNCBP2-AS2 #2, n = 122. (E) Sprouting quantification of ECs treated with the CM from hypoxic pCAFs transfected with siCTL or siNCBP2-AS2 (two different siRNAs) in the presence of bevacizumab or IgGK (control). N = beads assessed in three biological replicates. CM with IgG: siCTL, n = 121; siNCBP2-AS2 #1, n = 108; and siNCBP2-AS2 #2, n = 107. CM with bevacizumab: siCTL, n = 126; siNCBP2-AS2 #1, n = 130; and siNCBP2-AS2 #2, n = 110. (F) RT-qPCR showing the remaining mRNA levels of NCBP2-AS2 normalized to TBP. (G) Representative Western blot for proteins downstream of VEGFR signaling in ECs treated with the CM from cCAFs silenced or not for NCBP2-AS2. N = three biological replicates. (H) RT-qPCR showing remaining mRNA levels of NCBP2-AS2 normalized to TBP. (I) Working model. *P < 0.05, **P < 0.01, ****P < 0.0001, t test.

NCBP2-AS2 is expressed in human tumors

Last, we determined NCBP2-AS2 expression in publicly available The Cancer Genome Atlas (TCGA) datasets of breast cancer patient samples. While this analysis showed that NCBP2-AS2 is expressed in tumors, given that NCBP2-AS2 regulation by hypoxia occurs predominantly at the posttranscriptional level, we could not find a positive correlation between NCBP2-AS2 gene expression levels and hypoxia (fig. S9). We then looked at NCBP2-AS2 expression in a tissue microarray (TMA). Because antibodies that recognize NCBP2-AS2 by immunohistochemistry staining are not available, we assessed its expression by in situ hybridization. In a breast cancer TMA containing 108 patient samples, we observed NCBP2-AS2–positive staining in the stroma of 10% of the tumors. In particular, ductal carcinomas positively stained for NCBP2-AS2, whereas no staining was detected in the stroma of lobular carcinomas. We found that 3% of intraductal carcinoma and 14% of invasive ductal carcinoma positively stain for NCBP2-AS2 in the stroma (Fig. 8, A and B). These findings indicate that the frequency of NCBP2-AS2 expression in the stroma increases with the progression of the disease, suggesting a possible role in advanced disease.

Fig. 8 NCBP2-AS2 is expressed in the stroma of breast tumors.

(A) Pie chart showing, for each type of breast cancer contained in a breast cancer TMA of 100 patient samples, how many showed a positive staining (as assessed by in situ hybridization with RNAscope) for NCBP2-AS2 in the tumor stroma. (B) Representative images of breast cancer tissues with positive or negative staining (as assessed by in situ hybridization with RNAscope) for NCBP2-AS2 in the tumor stroma. Scale bar, 100 μm (left) and 10 μm (right).

DISCUSSION

The ability of hypoxia to create an aberrant and dysfunctional tumor vasculature is underpinned by secreted pro-angiogenic factors that are downstream of HIF1α signaling, of which VEGF plays a key role. CAFs are highly abundant in invasive carcinomas where they create an aberrant microenvironment permissive for blood vessel growth. Notably, hypoxic and pro-angiogenic gene expression signatures found in the stroma of breast cancer patients correlate with poor clinical outcome (47), highlighting the importance of better understanding the role of hypoxia in the tumor stroma. Here, we showed that hypoxia exacerbated the pro-angiogenic functions of CAFs, and MS-proteomic analyses pinpointed potential molecular mechanisms responsible for this function. VEGF is the major, but not the only, contributor of the enhanced pro-angiogenic activity of hypoxic CAFs, and we unveiled a mechanism through which NCBP2-AS2 controls VEGFA expression in hypoxic CAFs in vitro.

The CAF secretome is unique for its complexity and ability to remodel the tumor microenvironment including recruiting blood vessels. The cCAF model used in this study has been previously exploited by us and others to determine that transforming growth factor–β (TGFβ)– and CXCL12-dependent pathways support the pro-angiogenic activity of these cells (13) and to uncover a pro-angiogenic function for the CAF-secreted protein chloride intracellular channel 3 (CLIC3) (29). Here, we deployed cCAFs and pCAFs to draw an intracellular and extracellular molecular portrait of hypoxic mammary CAFs. We did not find significant changes in CAF activation marker αSMA (ACTA2 in data file S1), indicating that in our cell models and experimental conditions, hypoxia does not inactivate CAFs, as previously shown in head and neck and vulvar CAFs (18). Conversely, reduced oxygen availability remodels the CAF secretome. Hypoxic CAFs secrete increased amounts of pro-angiogenic factors and decreased amounts of anti-angiogenic factors, conferring enhanced pro-angiogenic properties to their CM. We asked whether our datasets could be exploited to identify mechanisms that promote the pro-angiogenic effect of hypoxic CAFs. These could be potential targets to block the formation of a dysfunctional vasculature in hypoxic cancers. Proteins up-regulated by hypoxia and which were found also secreted by CAFs may be useful not only as targets but also as potential biomarkers in body fluids. We discovered the antisense gene protein NCBP2-AS2 (transcribed from the antisense DNA strand of the gene NCBP2) as a hypoxia-responsive molecule. Hypoxia controls protein levels and regulates both transcription and translation mechanisms (48, 49). Notably, the latter promotes a swifter response to environmental stimuli. NCBP2-AS2 translation was enhanced by hypoxia, thus indicating that it may play key roles to counteract hypoxic stress in the cells, for example, boosting VEGFA signaling for the recruitment of blood vessels. We showed that targeting NCBP2-AS2 in hypoxic CAFs reduced VEGFA expression and secretion to the levels of VEGFA expressed in normoxic CAFs, but that this manipulation did not affect the pro-angiogenic activity of normoxic CAFs in our in vitro setup. These results make it tempting to speculate that NCBP2-AS2 could offer a therapeutic opportunity to normalize the dysfunctional vasculature typical of hypoxic malignant cancers without altering physiological angiogenesis. Silencing NCBP2-AS2 in hypoxic CAFs substantially reduced the pro-angiogenic activity of their CM in vivo. In situ hybridization for NCBP2-AS2 in breast cancer patient samples showed that it is expressed in the tumor stroma. This prompts investigation into the functional consequences of modulating the levels of stromal NCBP2-AS2 in cancer pathology in vivo.

How NCBP2-AS2 controls VEGFA levels remains to be elucidated. STC1 may play a role because it was also found down-regulated in the CAF secretome in the absence of NCBP2-AS2. In gastric cancer, STC1 promotes angiogenesis in vitro and in vivo by increasing VEGF expression through the activation of protein kinase C βII (PKCβII) and ERK1/2 pathways (35). Additional players are suggested by unbiased MS-based interactome analyses in HeLa cells (50), which identified the GTPase KRAS, the Ras-related protein Rab-5C (RAB5C), the nuclear cap-binding protein subunit 2 (NCBP2), the serine/threonine-protein phosphatase 2A regulatory subunit B″ subunit gamma (PPP2R3C), the kinesin-like protein KIF19 (KIF19), the RNA-binding protein 8A (RBM8A), and the survival motor neuron protein (SMN1) as potential binding partners of NCBP2-AS2. Of note, SMN1 and VEGF are up-regulated in ischemic rat hippocampus (51) and VEGF expression is up-regulated by wild-type KRAS in human colon carcinoma cells (52) and by mutant KRAS in human colon carcinoma (53), pancreatic epithelial (54), and NIH3T3 (55) cells.

In conclusion, here, we provide new insights into how mammary CAFs adapt to and behave in a low oxygen environment and provide a resource to discover molecular mechanisms underpinning CAF functions. Furthermore, we unraveled a mechanism through which hypoxic CAFs control the expression of the most potent pro-angiogenic factor, VEGF, through NCBP2-AS2, which we propose to rename to a more functional name, HIAR (hypoxia-induced angiogenesis regulator).

MATERIALS AND METHODS

Cell culture

Human umbilical vein ECs (HUVECs) were isolated in house from umbilical cords and cultured in EGM-2 medium (Lonza) on plastic dishes coated in 1% gelatin. pCAFs were isolated in-house from a grade 3 triple-negative breast cancer patient sample obtained through National Health Service (NHS) Greater Glasgow and Clyde Biorepository. All participants gave specific consent to use their tissue samples for research. The pCAFs were cultured on plastic dishes coated with collagen I (35 μg/ml; Gibco). The cCAF cell line was provided by Akira Orimo (Paterson Institute, Manchester). The pCAFs, cCAFs, and human embryonic kidney (HEK) 293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Life Technologies), 1% glutamine (Life Technologies), and 1% penicillin/streptomycin (Life Technologies). For quantitative proteomics experiments, pCAFs and cCAFs were cultured in SILAC DMEM (Life Technologies) supplemented with 2% FBS, 8% 10 kDa dialyzed FBS (PAA), 1% glutamine, and 1% penicillin/streptomycin. SILAC DMEM used for the “medium” labeled cells contained 13C6 l-arginine (84 mg/liter) and D4 l-lysine (175 mg/liter), whereas the medium for the “heavy” labeled cells contained 13C615N4 l-arginine (84 mg/liter) and 13C615N2 l-lysine (175 mg/liter; Cambridge Isotope Laboratories). All the cell lines described above were kept in standard tissue culture incubators at 37°C, 5% CO2 in 21% O2 (normoxic condition) or in a hypoxic chamber at 37°C, 5% CO2 in 1% O2 (hypoxic condition). The hypoxic chamber used was Whitley H30/H35 Hypoxystation (Don Whitley Scientific). All cell lines were routinely tested for mycoplasma.

pCAF isolation

Tumor tissue was minced and transferred into a 10-cm culture dish. Ten milliliters of RPMI supplemented with 10% fetal calf serum, 1% fungizone (Life Technologies), 1% penicillin/streptomycin (Invitrogen), and activated collagenase 1A (1 mg/ml, Sigma) was added and incubated overnight at 37°C. The digested tissue was then filtered through a cell strainer and centrifuged. The CAF-containing cell pellet was washed in phosphate-buffered saline (PBS); resuspended in DMEM, 10% FBS, 1% penicillin/streptomycin, and 1% fungizone; and plated on collagen-coated (35 μg/ml; rat tail collagen 1, Gibco) culture dish. Patient sample was obtained through NHS Greater Glasgow and Clyde Biorepository (project #305). All participants gave specific consent to use their tissue samples for research.

pCAF immortalization

The pCAFs were immortalized using a human telomerase reverse transcriptase (hTERT)–expressing plasmid (pIRES2-hygro), provided by Franco Calvo (Institute of Cancer Research, London). Lentivirus containing the hTERT plasmid was generated in HEK293T cells. Two rounds of viral transduction were performed on consecutive days. Cells were selected using hygromycin (10 μg/ml; Sigma).

3D fibrin gel–based sprouting assay

To assess the sprouting ability of ECs in 3D, we have used a previously described protocol (26) with minor modifications (29). Around 3.3 × 105 ECs embedded in the fibrin gel were either treated with CAF-derived CM (prepared from CAFs cultured in 2D on plastic dish) or cocultured with 2 × 104 CAFs in a 24-well plate. For the experiments where the CAF-derived CM was used, 900 μl of freshly collected CM derived from 8 × 104 CAFs cultured in EGM-2 for 72 hours under normoxia or hypoxia was admixed with 100 μl of fresh EGM-2. The CM was replaced with fresh CM every 24 hours. For experiments using VEGFA-neutralizing antibody, bevacizumab (1 μg/ml; Roche) or IgGK (Sigma) was added in the CM. For the rescue experiments, VEGFA (50 ng/ml; Sigma) was added. For the 3D experiments, the images used for quantification were acquired using the Axiovert 25 microscope (Zeiss) with a Retiga EXi Fast 1394 camera (Imaging) 2 days after the beads were embedded in the fibrin gel. Three technical replicates were performed for each biological replicate. A total of 15 images were acquired and quantified for each biological replicate. In total, n = 3 biological replicates (n = 45 images) were performed for each experiment.

EdU proliferation assay

EdU (Life Technologies) was added to the culture medium to a final concentration of 10 μM, and cells were allowed to incorporate for 6 hours, when cultured in 3D, and 1.5 hours, when cultured in 2D. Following EdU incorporation, cells were fixed with 4% paraformaldehyde (PFA) for 15 min, washed twice with 3% bovine serum albumin (BSA), and permeabilized with 0.5% Triton X-100 for 20 min. EdU-positive cells were fluorescently labeled using a click reaction following the manufactures’ instructions (Life Technologies), whereas all cells were stained with 4′,6-diamidino-2-phenylindole (DAPI). After the staining, images were acquired in the Nikon A1R confocal microscope for the 3D assay and in the Zeiss 710 confocal for 2D cultures. The percentage of cells that entered the S phase of the cell cycle was determined by dividing the number of EdU-positive cells by the total number of cells positive for DAPI.

TUNEL assay

Click-iT TUNEL Alexa Fluor 488 Imaging Assay (Life Technologies) was used to monitor DNA degradation in apoptotic cells. In a 24-well plate, 8 × 104 cells were seeded on a 13-mm coverslip. CAFs were exposed to hypoxia and fixed with 4% PFA for 15 min. Cells were permeabilized with 0.25% Triton X-100 in PBS for 20 min, and the TUNEL assay was performed according to the manufacturer’s instructions.

Secretome production and analysis

For the cCAF- and pCAF-derived secretome, 6 × 106 medium and heavy SILAC-labeled cells were seeded in 15-cm dishes. The following day, cells were washed six times with prewarmed PBS to remove FBS residuals, serum-free SILAC medium was added, and cells were allowed to secrete for 72 hours in either normoxic or hypoxic conditions. Subsequently, the same volume of CM from three dishes from each condition was mixed and subjected to sequential centrifugations at 4°C: 300g for 10 min, 2000g for 10 min, and 10,000g for 30 min. An ultracentrifugation step of 100,000g for 70 min was carried out to separate the soluble secretome from the EV fraction. The pellet containing the EVs was lysed with 8 M urea buffer and in-solution–digested with trypsin (Promega). The soluble fraction was acidified using 10% trifluoroacetic acid (TFA) to pH 5.0, and proteins were enriched using StrataClean Resin (Agilent Technologies) as previously described (29). The eluted proteins from the resin were loaded in a 4 to 12% gradient NuPAGE Novex Bis-Tris gel, stained with instant blue (Expedeon), and in-gel–digested (56) with trypsin (Promega). For the secretome of hypoxic cCAFs silenced for NCBP2-AS2, medium and heavy SILAC-labeled cells were transfected with siRNA CTL and siRNA NCBP2-AS2. Cells (2 × 106) from each condition were seeded in a 10-cm dish, and the CM was prepared as described above. In this experiment, all secreted proteins and EVs were enriched with StrataClean Resin and in-gel–digested with trypsin.

Total proteome sample preparation

For the analysis of the total proteome, total lysates from the same cells used for the isolation of the secretome were collected using 2% SDS, 100 mM tris-HCl (pH 7.4) lysis buffer. Proteins were digested into peptides using filter-aided sample preparation using a filter with a 30-kDa cutoff [FASP; (57)]. For the proteomes of cCAFs and pCAFs exposed to hypoxia, 50 μg of FASP-derived peptides was fractionated using strong anion exchange chromatography [SAX; (30)]. For the proteome of hypoxic cCAFs silenced for NCBP2-AS2, 50 μg of FASP-derived peptides was fractionated using high-pH reverse-phase on tip fractionation as previously described (58), with minor modifications.

MS analysis

After trypsin digestion, peptides were desalted using StageTip (59). Peptides were resuspended in 1% TFA, 0.2% acetic acid, or formic acid buffer and injected on an EASY-nLC or 1200 system coupled online to an LTQ-Orbitrap Pro Elite or Q-Exactive HF, respectively, via nanoelectrospray ion source (Thermo Fisher Scientific). Peptides were separated on a 20-cm fused silica emitter (New Objective) packed in-house with reverse-phase Reprosil Pur Basic 1.9 μm (Dr. Maisch GmbH). For peptide elution, buffer A (0.1% formic acid) and buffer B (80% acetonitrile, 0.1% formic acid) were used. In the Orbitrap Elite, peptides were eluted with a flow of 200 nl/min from 5 to 30% of buffer B, in a 90-min linear gradient (for in-gel–digested peptides) and 190-min gradient (for SAX fractions and EVs). In the Q-Exactive HF, peptides were eluted with a flow of 300 nl/min from 5% to 30% of buffer B in a 60-min linear gradient. In the Orbitrap Elite, the mass range acquired for the full MS scan was 300 to 1650 m/z, with a resolution of 120,000 at 400 Th. The top 10 most intense ions in the full MS were isolated for higher-energy collision dissociation fragmentation with a target of 40,000 ions at a resolution of 15,000 at 400 Th. For the Q-Exactive HF, the mass range acquired for the full MS scan was 375 to 1500 m/z, with a resolution of 60,000 at 250 Th. The top 10 most intense ions in the full MS were isolated for fragmentation with a target of 50,000 ions at a resolution of 15,000 at 250 Th. MS data were acquired using the XCalibur software (Thermo Fisher Scientific).

Parallel reaction monitoring and data analysis

The mass spectrometer was operated in positive ion mode and used in parallel reaction monitoring (PRM). A data-dependent experiment with inclusion list was set up together with a full scan. The full scan was performed at a target value of 1 × 106 ions with resolution of 60,000 at m/z 200 over mass range of 375 to 1600 m/z followed by six PRM scans. The inclusion mass list included light or medium and heavy SILAC-labeled NCBP2-AS2 peptides 5 to 18, 70 to 77, and 82 to 92. Targeted peptides were constantly fragmented in the higher-energy collisional dissociation (HCD) cell (Q-Exactive HF) or linear trap (Orbitrap Elite) and acquired into the Orbitrap analyzer or ion trap, respectively, at a resolution of 15,000 at 200 m/z (Q-Exactive HF). Quadrupole isolation width was set to 1.2 m/z (Q-Exactive HF) or 1 m/z (Elite), and normalized collision energy of 28 (Q-Exactive HF) or 35 (Orbitrap Elite) was used to fragment a maximum target of 1 × 105 ions (or injection time of 25 ms for the Q-Exactive HF) or 7.5 × 103 ions in the Elite.

The PRM data traces of NCBP2-AS2 tryptic peptides were used for MS/MS-based quantification. Raw data were imported into Skyline (version 4.1.0.18169), and y and b product ion chromatograms were extracted, summed, and integrated. Only fragment ion peaks with the same retention time were used for quantification. In the transition settings menu, the “MS/MS filtering tool” was applied to all replicates using “targeted” acquisition method. The product mass analyzer was set to “Orbitrap” and resolution to 15,000 (Q-Exactive HF) or to “QIT” with a tolerance of 0.5 m/z (Elite). All matching scans were included in data extraction. Ion series extracted were compared to the five most intense product ions of an NCBP2-AS2 SILAC medium and heavy peptide library built with previously acquired data. For the siRNA, .raw files were analyzed with MaxQuant; protein quantification was according to the MaxQuant label-free quantification algorithm.

MS data analysis

The MS .raw files were processed with the MaxQuant software (31) version 1.5.0.36 and searched with the Andromeda search engine (60) with the following settings: a minimal peptide length of 7 amino acids; a multiplicity of 2, where the light labels were Arg6 and Lys4 and the heavy labels were Arg10 and Lys8; variable modifications acetyl (protein N-term) and oxidation (M); fixed modification carbamidomethyl (C); specificity for trypsin cleavage was required; and a maximum of two missed cleavages were allowed. For the search of the parent mass and fragment ions, an initial mass deviation of 4.5 and 20 ppm, respectively, was required. The re-quantify parameter was enabled, and match between runs was used between similar fractions. For identification, the false discovery rates at the protein and peptide level were set to 1%. For accurate protein quantification, only unique peptides with at least two ratio counts were used. A ratio count of 1 was used for the total proteome of CAFs silenced for NCBP2-AS2 to allow the quantification of NCBP2-AS2. The MaxQuant output table “ProteinGroups.txt” was uploaded in Perseus (version 1.5.2.11) (61) for downstream analysis. The data were filtered to remove potential contaminants, reverse peptides that match a decoy database, and proteins only identified by site, which corresponds to proteins identified through peptides that were identified only in their modified version. To assure an unambiguous identification, only proteins identified with at least one unique peptide were considered. The normalized SILAC ratio was used for the analysis. Ratios from the “reverse” experiment were inverted. Then, ratios and intensities were logarithmized, log2 and log10, respectively. When the distribution of the SILAC ratios was not centered on the zero, data were normalized by subtracting the median of the all SILAC ratios calculated for that experiment. Annotation from Gene Ontology Cellular Compartment (GOCC), Gene Ontology Biological Processes (GOBP), Gene Ontology Molecular Function (GOMF), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Protein families (Pfam) databases were added. For the STRING analysis, the list of the regulated proteins was uploaded in STRING version 10.0 (62) and the analysis was conducted using the default settings. The output files were visualized with Cystoscape (version 3.4.0) (63).

Knockdown with siRNA

cCAFs and pCAFs (90% confluent) were harvested, and 2 × 106 cells were used in each transfection according to the manufacturer’s protocol using the program T-20 and the Amaxa kit R (Lonza). Cells were transfected with 1 to 3 nM of the nontargeting siRNA as a control (D-001810-10-05, GE Healthcare Dharmacon) or with siRNAs targeting NCBP2-AS2 (NCBP2-AS2 #1, CTGGTATAGCAGGACAGTTAA; NCBP2-AS2 #2, CAGGCAAGATGTGTTAAGAAT) (Qiagen). Cells were used for experiments 24 hours after silencing.

Scratch wound migration assay

HUVECs (6 × 104) were seeded in each well of an image lock 96-well plate (Essen Bioscience). The following day, wounds were made using the wound maker (Essen Bioscience) and the CM derived from CAFs silenced or not for NCBP2-AS2 was added. The images were acquired every 30 min for 24 hours using Incucyte Zoom (Essen Bioscience). Relative wound density over the time was measured using the Incucyte Zoom software (Essen Bioscience).

Protrusions and lamellipodia formation assay

HUVECs (1 × 104) were seeded on a 13-mm coverslip coated with gelatin. The following day, cells were starved for 4 hours in EBM2. Then, cells were treated for 1 hour with CM derived from hypoxic CAFs siCTL, siNCBP2-AS2 #1, or siNCBP2-AS2 #2. Cells were fixed in 4% PFA for 10 min, permeabilized with 0.2% Triton X-100 in PBS for 5 min, blocked in 5% BSA for 5 min, and incubated with anti-cortactin antibody for 1 hour (1:200, Millipore 4F11). For F-actin staining, we used Alexa Fluor 594 phalloidin (1:200, Molecular Probes, A12381). For nuclei staining, DAPI was used (1:1000, Sigma). Images were acquired on a Nikon A1R confocal microscope.

EC treatment for VEGF signaling assessment

HUVECs (2 × 106) were seeded on a gelatin-coated 15-cm dish. The following day, cells were starved for 4 hours in EBM2 and treated for 5 min with the CM derived from hypoxic CAFs siCTL, siNCBP2-AS2 #1, or siNCBP2-AS2 #2. As a negative control for VEGF signaling, we used cells starved in EBM2; as a positive control, following 4-hour starvation, cells were treated for 5 min with VEGFA (20 ng/ml; Sigma).

Enzyme-linked immunosorbent assay

VEGFA concentration in the CM was measured using the Quantikine ELISA Human VEGF Kit (DVE00, R&D Systems) according to the manufacturer’s instructions. Each sample was analyzed in three technical replicates.

Swap CM to assess VEGF levels

The cCAFs were transfected with NCBP2-AS2 #1 or control siRNA as described above. The CM was derived from 8 × 104 transfected cells cultured for 72 hours under hypoxia. After centrifugation, the cleared CM was used to culture new NCBP2-AS2 #1 siRNA-transfected cCAFs. After 72-hour culture under hypoxia, VEGFA levels in the CM were measured by ELISA (see above). VEGFA secretion by the medium-conditioning cells was accounted for by incubating the CM in adjacent empty wells during the experiment and subtracted from the ELISA values to give the results shown.

Western blotting analysis

Total lysates were obtained using a lysis buffer containing 2% SDS, 100 mM tris-HCl (pH 7.4), incubated at 95°C for 5 min, sonicated using a metal tip (Soniprep 150, MSE), and centrifuged at 16,000g for 10 min. Protein concentration was determined using Optiblot Bradford reagent (Abcam). Twenty micrograms of proteins were separated using 4 to 12% gradient NuPAGE Novex Bis-Tris gel (Life Technologies). Protein transfer was performed on methanol-activated polyvinylidene difluoride or nitrocellulose membrane. Membrane was blocked overnight in 5% nonfat milk (Marvel) or in 3% BSA (Sigma) in TBST (tris-buffered saline polysorbate 20) at 4°C. The following primary antibodies were used: PTX (1:500, 13797-1-AP, ProteinTech), CPA4 (1:500, 26824-1-AP, ProteinTech), HIF1α (1:500, 10006421, Cayman), vinculin (1:2000, V9131, Sigma), glyceraldehyde-3-phosphate dehydrogenase (1:1000, Santa Cruz Biotechnology, sc48167), total VEGFR2 [1:2000, 55B11, Cell Signaling Technology (CST)], phospho-VEGFR2 (Tyr1175) (1:1000, 2478S, CST), total PLCγ (1:2000, 2822S, CST), phospho-PLCγ (Tyr783) (1:1000, 2821S, CST), total ERK1/2 [p44/42 mitogen-activated protein kinase (MAPK)] (1:2000, 9102, CST), phospho-ERK1/2 (p44/42 MAPK—Thr202/Tyr204) (1:1000, 9106, CST), total p38 MAPK (1:2000, 8690S, CST), phospho-p38 MAPK (Thr180/Tyr182), (1:2000, 9216S, CST), total AKT (1:1000, D7D11, CST), and phospho-AKT (Ser473) (1:1000, 9271, CST). LI-COR secondary antibodies (1:10,000, LI-COR Biosciences) or horseradish peroxidase–conjugated antibodies (1:5000, New England Biolabs) were incubated for 1 hour at room temperature. Western blot images were acquired using Odyssey CLx Scanner (LI-COR Biosciences) or myECL Imager (Thermo Fisher Scientific).

Reverse transcription polymerase chain reaction

The primers for RT-qPCR used in this study were designed using Primer Blast (National Center for Biotechnology Information): NCBP2-AS2 [GCGCTGAGCGATTTAGAGAC (forward) and ACTGTAGAGCCCGGAAGGAA (reverse)], VEGFA [GCTACTGCCATCCAATCGAG (forward) and CATCGCATCAGGGGCACAC (reverse)], STC1 [TCTGGTGCTGGTGATCAGTG (forward) and CACTTCAGCTGAGTTTTGAGCC (reverse)], CPA4 [ACCTGCAGGCCCTTTTAGACA (forward) and AGGAAAGTCTGCGGCAATGT (reverse)], LOXL2 [CCCCCTGGAGACTACCTGTT (forward) and TCGCTGAAGGAACCACCTAT (reverse)], and PTX3 [GCCGGCAGGTTGTGAAACAG (forward) and ACATCTGTGGCTTTGACCCAA (reverse)]. Tata-binding protein [TBP; AGTGACCCAGCATCACTGTTT (forward) and TAAGGTGGCAGGCTGTTGTT (reverse)] and 18S [AGGAATTGACGGAAGGGCA (forward) and GGACATCTAAGGGCATCACA (reverse)] were used as housekeeping genes for normalization. For polysomal profiling, the following primers were used: Human β-actin CTCTTCCAGCCTTCCTTCCT (forward) and CACCTTCACCGTTCCAGTTT (reverse) and the NCBP2-AS2 primers listed above. For mRNA isolation, 5 × 105 cells were seeded in a six-well plate. After hypoxia (1% O2, 72 hours) or DMOG (1 mM, 24 hours) treatment, mRNA was collected using RNeasy Lysis Buffer (RLT) buffer (Qiagen). Total RNA extraction and deoxyribonuclease treatment were performed following the manufacturer’s instructions of the RNeasy kit (Qiagen). Complementary DNA (cDNA) was synthesized using an iScript kit (Bio-Rad) using 500 ng to 1 μg of RNA. cDNA was diluted 1:10, and 2 μl was used in each RT-qPCR reaction (technical triplicates), with 10 μl of iTAQ Universal SYBR Green Supermix (Bio-Rad) and 400 nM of forward and reverse primers diluted in 8 μl of water. Each plate contained the appropriate nontemplate controls. Runs were performed in 7500 Fast Real-Time PCR System (Applied Biosystems).

Polysomal RNA isolation

Cells exposed to normoxia or hypoxia for 24 hours were treated with cycloheximide (100 μg/ml) for 3 min before harvesting. Cell pellets were lysed in lysis buffer [300 mM NaCl, 15 mM MgCl2, and 15 mM tris-HCl (pH 7.5), heparin sulfate (1 mg/ml), cycloheximide (100 μg/ml), and 0.1% Triton X-100]. Post-nuclear lysates were layered on 10 ml of 10 to 50% (w/v) sucrose gradients of gradient buffer (lysis buffer without Triton X-100). Gradients were centrifuged at 256,824g for 2 hours at 4°C in a SW40Ti rotor (Beckman Coulter) and separated through a live absorbance 254-nm ultraviolet spectrometer (Isco). Gradients were separated into 10 fractions, and RNA was precipitated by addition of guanidine hydrochloride (2.9 M final concentration) and ethanol (50%, v/v). Precipitated RNA was then ethanol-precipitated for a second time before use. Data are represented as total RNA quantity in fractions from nontranslating and polysomal sections of the gradients.

Sponge angiogenesis assay

All mouse procedures were in accordance with ethical approval from the Institutional Animal Care and Research Advisory Committee of the K.U. Leuven. We used the subcutaneous sponge assay described in detail in the JoVe protocol (44), with minor modifications. A biopsy puncher was used to obtain 6-mm-diameter sponges (Haromed, 1255889) for subcutaneous injection. Each sponge was embedded with 25 μl of CM produced for 48 hours by hypoxic CAF siCTL or siNCBP2-AS2. Four sponges per animal were subcutaneously transplanted and surgically dissected after 7 days. Each sponge was cut in two halves and embedded in octadecylamine. Frozen sections of each sponge were stained for blood vessels. Meca32 (1:100, BD Biosciences, 550563) was used for the ECs and Ng2 (1:100, Abcam, AB5320) for the pericytes. Whole sections were imaged as a tile scan and analyzed by a macro used in ImageJ. The channels were split before the individual images were put together into individual mosaic images for each channel and subsequently merged. Blood vessels containing regions were used for the analysis. Meca32 colocalization with Ng2 within the zone was measured.

TMA staining

In situ hybridization staining to detect mRNA expression in tissue sections was performed on tumors formalin-fixed for 24 hours at room temperature. The probe to detect the human NCBP2-AS2 was designed by Advanced Cell Diagnostics (Hs-NCBP2-AS2 gene 300031 ID: 152217). The staining of the paraffin-embedded sections was performed following the manufacturer’s instructions (Advanced Cell Diagnostics). Breast cancer TMA was purchased and contained single core per case (BR2082b, Insight Biotechnology). One hundred eight cases were analyzed, which include invasive ductal carcinoma, lobular carcinoma, intraductal carcinoma, and normal tissues. Cores at the border of the slide were not included in the analysis because they were damaged. Cores were defined positive for NCBP2-AS2 staining if positive cells were detected in the stromal compartment. Slides and TMA were digitized and viewed using Slidepath Digital Image Hub (Leica Microsystems) and quantified using Halo software (ACD).

NCBP2-AS2 recombinant protein

Human NCBP2-AS2 was cloned in Bam HI/Eco RI of pGEX-4T-1 vector. NCBP2-AS2 was expressed as GST fusion protein in BL21 Codon-plus (DE3). After overnight induction at 18°C, cells were lysed and the protein was purified using a two-step chromatography procedure, including a GST affinity and size exclusion chromatography. Protein was aliquoted and stored in a buffer containing 150 mM NaCl, 20 mM tris (pH 7.5), and 2 mM dithiothreitol.

Statistical analysis

GraphPad Prism version 7.0 was used for statistical analyses. For the quantification of the EC sprouting, first, the distribution of the data was evaluated using the D’Agostino and Pearson normality test. For data with normal distribution, an unpaired two-tailed t test with Welch’s correction was used. When the distribution was not normal, the Mann-Whitney t test was used. For other experiments, the two-tailed unpaired t test with Welch’s correction was used to compare samples. For cases in which samples were normalized to a control group, the normalized data were first log-transformed and then a one-sample t test (against a comparison value of 0) was performed on the resulting log-transformed data. For the cortactin quantification, Kruskal-Wallis test with Dunn’s multiple comparison test was performed. P ≤ 0.05 was considered statistically significant. Data are represented as mean ± SEM, unless stated otherwise. Data in the plot were obtained from three biological replicates, unless stated otherwise. For the MS proteomic data analysis, the Perseus software was used: We required proteins to be quantified in at least two of three biological replicates. Then, the SILAC ratios of these quantified proteins were averaged by the mean. For the cCAF and pCAF hypoxia-induced proteome and secretome, the threshold to define statistically significantly up-regulated and down-regulated proteins was defined to be higher than one and two SDs from the mean of the SILAC ratios. This relaxed statistical threshold was chosen because known HIF1α-responsive genes were found regulated within this threshold. For the NCBP2-AS2 knockdown experiment, one-sample t test analysis was performed to ease visualization of the results with volcano plot.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/12/567/eaan8247/DC1

Fig. S1. Response of CAFs to hypoxia.

Fig. S2. Hypoxia remodels the secretome of CAFs to promote angiogenesis.

Fig. S3. Hypoxia-responsive proteins discovered in our study.

Fig. S4. NCBP2-AS2 identification and quantification.

Fig. S5. NCBP2-AS2 regulation by hypoxia.

Fig. S6. PRM of NCBP2-AS2 in pulsed-SILAC and siRNA experiment.

Fig. S7. Extracellular NCBP2-AS2 does not induce sprouting in ECs and VEGF expression in CAFs.

Fig. S8. Western blot quantification of VEGFR signaling in HUVECs.

Fig. S9. NCBP2-AS2 gene expression levels do not correlate with hypoxia in tumors.

Data file S1. Proteome and secretome analysis of mammary CAFs.

Data file S2. Hypoxia-regulated proteins in CAFs.

Data file S3. Proteome of CAFs expressing siNCBP2-AS2.

Data file S4. Secretome of CAFs expressing siNCBP2-AS2.

REFERENCES AND NOTES

Acknowledgments: We thank C. Nixon, the BSU, the BAIR, and the histology facilities at the CRUK Beatson Institute; C. Orange for histopathology services; NHS Greater Glasgow and Clyde Biorepository for providing patient samples; and the PRIDE team. Funding: This work was funded by Cancer Research UK [CRUK Beatson Institute A17196 (to F.G.K., S.J.A., L.J.N., S.L., D.M.B., A.J., and E.K.M.); CRUK Glasgow Centre A18076 and Stand Up to Cancer campaign for Cancer Research UK A12935 (to S.Z.), A19257 (to S.I.), A15673 (to L.M.M.), and A21139 (to L.M.M., J.R.P.K., and O.J.S.); and CRUK Science Committee Programme Award A24388 (to O.J.S.)] and ERC Starting Grant ColonCan agreement 311301 (to J.R.P.K.). M.M. and J.S. were funded by FWO (G0D1717N) and Stichting tegen Kanker. Author contributions: F.G.K. and S.Z. conceptualized the study. F.G.K., J.R.P.K., A.J., and S.Z. developed methodology. F.G.K., L.J.N., S.J.A., J.S., A.J., S.L., and J.R.P.K. performed investigation. S.I. and D.M.B. provided resources. F.G.K., J.R.P.K., A.J., E.K.M., S.L., and S.Z. performed data curation and formal analysis. F.G.K. and S.Z. wrote the original draft. F.G.K. and S.Z. visualized the data. S.I., O.J.S., L.M.M., M.M., and S.Z. supervised the work. Competing interests: O.J.S. has received funding from Astra Zeneca, Novartis, Jansen, and Cancer Research Technologies. All other authors declare that they have no competing interests. Data and materials availability: The .raw MS files and search/identification files obtained with MaxQuant have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org/cgi/GetDataset) via the PRIDE partner repository (64) with dataset identifier PXD006535s (hypoxia datasets) and PXD006539 (siNCBP2-AS2 datasets). All other data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.
View Abstract

Stay Connected to Science Signaling

Navigate This Article